That quick ChatGPT draft of a warning letter may end up in court

Fast and convenient and it feels private, unfortunately, it might not be.

Fast and convenient and it feels private, unfortunately, it might not be.

If you’re a business owner or leader managing a team without dedicated HR support, you’ve probably done this when a tricky situation comes up, you’re short on time, so you jot the details into ChatGPT or Claude and ask it to help you draft the warning letter, the investigation notes or the performance review.

Fast and convenient and it feels private, unfortunately it might not be.

A recent US federal court case just tested exactly this question and while it’s about a criminal fraud trial rather than an HR matter, the reasoning behind it lands squarely on the desk of anyone drafting employee documents with AI.

What happened

In February 2026, a US court dealt with a case called United States v. Heppner. The defendant, facing fraud charges, had used Claude to help him think through his defence, then shared what he generated with his lawyers. He argued those AI chats should be protected by legal privilege, the same protection that keeps conversations between you and your lawyer confidential.

The judge disagreed, and for 3 reasons that translate directly to HR:

1. The AI isn’t a professional. Privilege exists because you’re talking to someone who holds a licence, has ethical obligations and can be disciplined if they breach your trust. An AI chatbot is none of those things, no matter how competent the response sounds.

2. It’s not confidential just because it feels private. Public AI tools generally collect what you type, use it to improve their models and can hand data over to regulators or in legal disputes. The moment you type something in, you’ve shared it with a third party.

3. Getting advice from AI isn’t the same as getting advice from a professional acting on your behalf. Even though the material was later shared with his actual lawyer, that didn’t retroactively make the AI chat itself protected.

Why this matters for the letters and files sitting in your inbox

Swap criminal defence for workplace investigation and the parallel is uncomfortably close. Think about what gets typed into public AI tools in a lot of businesses right now:

  • Disciplinary or termination letters, with real names and real allegations
  • Workplace investigation notes and interview summaries
  • Performance management documentation
  • Redundancy selection criteria and reasoning

If any of that ends up in a Fair Work Commission dispute, general protections claim or discrimination complaint, the prompts, drafts and AI outputs could be discoverable, meaning the other side, or the Commission, may be entitled to see them. Typing sensitive employee information into a free AI tool doesn’t create a protected, confidential record. It can do the opposite.

This isn’t a hypothetical lesson from the US either. Australian legal bodies are already onto it. The Law Society of NSW, the Legal Practice Board of WA, and the Victorian Legal Services Board have jointly warned that confidential or sensitive information shouldn’t go into public AI chatbots at all, because there’s no guarantee it stays secure. Australian courts apply the same core test as the US: privilege requires an actual lawyer, and an AI simply doesn’t qualify.

The Fair Work Commission is also seeing the fallout of AI use firsthand. Employment lawyers have reported cases of self-represented workers building entire unfair dismissal claims around AI chat advice, sometimes citing legal arguments that don’t even apply to their jurisdiction. The Commission’s response hasn’t been to ban AI use, but to make clear that whoever files something is still responsible for it being accurate. The same logic applies from the employer’s side: using AI to help you write something doesn’t hand off your responsibility for what goes in it, or protect what you typed from being seen later.

What this means for your business

You don’t need to swear off AI tools. You do need to be intentional about what goes into them.

  • Keep names, allegations and specific case details out of free, public AI tools
  • If you’re using AI to help structure a difficult letter, work from de-identified or hypothetical details, then add the real specifics yourself afterwards
  • Anything genuinely sensitive, an investigation, a termination, a legal grey area, should go through a qualified HR or legal adviser who can actually give you protected, professional advice
  • Don’t assume “I only used it to help me phrase something” will hold up if that draft later surfaces in a dispute

AI can absolutely help you work faster on HR documentation. Just don’t mistake speed for protection. If you’re not sure whether something needs a real second set of professional eyes before it becomes a permanent, discoverable record, that’s exactly the kind of call worth making with someone who can give you privileged advice, not just a fast draft.

Eve Nelson Managing Director, Red Bridge HR 0401 785 529 eve@redbridgehr.com.au


This post is general information and doesn’t constitute legal advice. Laws and regulations around AI and privilege are moving fast and vary by jurisdiction. Always check current sources and consult with a qualified employment lawyer about your specific situation.

Sources & further reading

Explore similar posts from us

With you every step of the way

Join our mailing list for advice & updates

Request a free intro call

Get started by telling us a little bit about your business.

Please note this is a consulting service for business owners and management only. This is not a support service support for individual employees.